Schneier on Security

Skeptical • October 6, 2014 8:28 PM

This is a powerful essay, and I appreciate the particular evidence that it cites in support of its arguments.

The question of the tradeoffs we face were communication devices immune to even lawful warrants is more directly confronted here than I’ve read previously – and it’s that feature which I appreciate the most.

On the one hand, there’s a strong argument made that, at least so far as court-ordered wiretaps are concerned, encryption has not foiled many government investigations.

On the other hand, I think the argument that imposing access requirements for lawful surveillance will necessarily render us quite vulnerable is weak.

Somewhat ironically, I see somewhat of a mirror between the rhetoric used by law enforcement to justify tighter restrictions on encryption, and the rhetoric used here (and elsewhere) to justify less tolerance for access mechanisms designed for law enforcement.

For instance,

Backdoor access built for the good guys is routinely used by the bad guys.

“Routinely” is a very strong claim. But the cases you cite are highly unusual. No statistics are adduced to demonstrate that access mechanisms designed to be restricted to government actors are often, or frequently, or routinely, exploited by the bad guys.

In this respect, that part of the argument in the essay resembles the anecdotes law enforcement will give to justify tighter restrictions on encryption.

And then there’s this claim:

You’re either vulnerable to eavesdropping by any of them, or you’re secure from eavesdropping from all of them.

Plug some other threat/security compliments into that framework.

“You’re either vulnerable to being kidnapped by anyone, or you’re secure from kidnapping by all of them.”

“You’re either vulnerable to being shot by anyone, or you’re secure from being shot by all of them.”

“You’re either vulnerable to anyone hijacking an aircraft, or you’re secure from hijackings against everyone.”

These all seem plainly unsound, as does the original formulation. But I then I thought that perhaps there was a meaning to “vulnerability” that I’m not taking into account – perhaps it has usage as a term of art. So, I glanced at a few sources, and was pleasantly surprised by Wikipedia’s entry on vulnerability.

I did not see a definition that would save the formulation.

But maybe I’m missing something.

How are you defining “vulnerability” in the above?

Is “vulnerability” a binary characteristic, in that one either is vulnerable to X or one is not, or are there degrees of vulnerability?

Why does a vulnerability to one threat imply vulnerability to all threats?

Skeptical • October 6, 2014 10:18 PM

@Daniel: First, I don’t think you know what the word example means. By definition an example is an illustration and not intended to be conclusive evidence.

I appreciate the heads up on what the word example can mean, but the word “example” wasn’t used in the portion of the essay we’re talking about. That’s not to detract from the value of the lesson, of course.

Your claim that Bruce’s evidence is based upon “highly unusual” cases is itself not backed up by any evidence. Where is your data?

As far as reporting goes, they’re pretty unusual. I’m not going to run a search over Lexis for you and post the results to prove the point though – and in fact, for the sake of argument, I’ll leave aside any claim about whether the cases provided are unusual or not. The burden of proof here is on Bruce.

As I understand it Bruce’s point is that criminal can and in fact do exploit legal means to get to their objectives–that seems unassailable to me.

Then you’ve missed his point. He said that bad actors routinely exploit mechanisms designed to permit only authorized law enforcement to use.

Routinely matters because we are weighing tradeoffs. If criminals routinely break mechanisms designed to permit only law enforcement, then clearly these mechanisms carry enormous cost to our security. However, if those mechanisms are brokenly rarely, and in some cases extremely rarely, then the cost to our security is much lower.

Again, your examples disprove your point. Bruce is not making a generalized security claim but a specific claim about the nature of the internet. The internet isn’t like kidnapping, or shooting, or hijacking an aircraft. The internet is the internet and by definition of the term “inter-net” it is a communal endeavor. Kidnapping, shooting, and hijacking are not communal endeavors.

Okay, so now we have:
1) The internet is the internet.
2) ???
3) Any vulnerability (in a system connected to the internet?) to anyone is a vulnerability to everyone.

Were Iranian centrifuges equally vulnerable to everyone? Greenpeace? Were Soviet undersea cables in certain areas equally vulnerable to everyone – or just to nations with incredibly advanced submarine technology?

Or do Iranian centrifuges and Soviet undersea cables not count because they’re not closely enough tied to the internet?

How about features on telecommunications equipment that enable government to effect a lawful wiretap? Vulnerable to everyone, including low-level drug dealers and bored folks surfing the internet looking for something interesting to listen to? How about script kiddies? How about foreign governments?

The only way for me to make sense of this argument is to give “vulnerability” a meaning that is almost completely divorced from the risk of exploitation that it presents.

So that law enforcement access on telecom equipment might constitute a vulnerability, but with very low risk of exploitation.

But then it’s not a question of whether we’re vulnerable, but rather a question of what risk we’re running.

Skeptical • October 7, 2014 8:29 AM

@Daniel: I don’t think the burden of proof is on Bruce. Bruce has made a claim and then provides some examples as evidence to back up that claim.

So now the examples are not simply illustrations, but should be viewed as evidence for his claim?

Then the problem remains that they’re no more sufficient than any other anecdote – they don’t persuade me that “bad guys” are routinely accessing, e.g. telecommunications equipment designed to facilitate access only for authorized access by the government.

If someone wants to contradict that claim the burden of proof is on them.

No, the burden of persuasion is on the individual who introduces the positive claim. In this case, that’s the claim I just described above. The examples aren’t sufficient evidence for the truth of the claim.

However, one of the major problem with deciding the claim–viz., whether it’s routine or not–is that hacking by definition is an underground activity. I make this point all the time in another context. Crime statistics don’t measure the actual incidence of crime. Crime statistics measure only the crime that got reported. And legal cases only measure crimes that got reported and where the person was caught. It may be that hackers using legal backdoors is in fact routine but they rarely get caught doing so. So I’m not sure that statistics are useful in solving the debate.

Sure, it could be something difficult to show.

As for the nature of the internet itself, you are once again confusing the ability to exploit with the fact of exploit. The fact that the Iranian centrifuges were not equally vulnerable to everyone is not material. The fact is that they were vulnerable. Whether that vulnerability is more or less “risky” is also not material because risk is always a matter of subjective opinion. In fact, one the things I appreciate most about Bruce’s essay is that it doesn’t go down the rabbit hole of “risk”. Everyone is vulnerable or no one is–risk has nothing to do with it.

I strongly disagree. Because we are deciding between tradeoffs, i.e. the benefits of having government be able to access communications with proper authorization and the costs of ensuring that access via a feature enabling it, the risk posed by that feature is very much central to the issue.

Skeptical • October 8, 2014 12:03 AM

@QnJ1Y2U: The problem, as Clive notes, is the nature of information objects and the internet expands the number of people with the resources to exploit a vulnerability by several orders of magnitude.

Well, Clive put forward the classic set of conditions under which a given encryption might be tested.

However, that classic set is not always applicable.

For an example, let’s suppose that Apple creates a key-escrow system for the iPhone, and then builds a database of ‘golden keys’ that can be used to decrypt those iPhones. They are now vulnerable to:

– Anyone with a court order

Given your next example, I assume here you mean a valid court order. The answer is no, unless you include the support of the personnel and means needed to authenticate and effect that order. But in any case, this is a misuse of the term vulnerability, unless it’s proper to say that my email accounts are vulnerable to being accessed by an authorized user with the proper credentials.

– Anyone who can fake a court order

No, because the order itself isn’t enough. There are procedures and means of effecting the order and authenticating it – so one would need to fake all of those as well.

– Anyone whose job is to execute the court orders

Is your picture of how this works that of a guy with a piece of paper who walks up to a technician, shows him the paper, and then waits for his connection?

– Anyone that Apple decides can have access outside a court order (NSLs)

NSLs aren’t a good example, but that’s extraneous to your point. As to whether Apple would have the capability to unlock unilaterally, that’s true of particular designs, but it’s not true of all designs.

– Anyone at an agency that gets a copy of the database (NSA)

The database of keys? This is a possibility under certain types of designs. Other designs might include features that require factors additional to the key.

– Any sysadmin at Apple who can access the database

And if there were none who could do so without additional authorization? And if the particular key obtained could only be used once other factors were satisfied?

– Anyone at the factory in China who is feeding the golden keys into the database

Seems easy to avoid.

– Anyone with access via a FOREIGN GOVERNMENT which requires Apple to supply them the golden keys as a requirement for entry into the their market

If the keys were designed in such a fashion, and if Apple complied, yes.

– Anyone that can bribe or extort anyone else in the list

Making lots of assumptions about the design of the system in question, yes.

And so on. And thanks to the expanding nature of attacks, leaks, and laws, the list will grow over time.

It’s a list highly contingent on the specifics of a given system.

The kind of list you’re producing here consists of a series of potential single points of failure for a given security structure. However, the structure need not be designed in such a fashion, and the resources needed to place any one of those points in jeopardy can be placed well out of the reach of most.

So my point again is simply that the risk posed by having authorized access features depends on the design of the system in which the feature exists.

Indeed, moreover and equally importantly, the risk depends on the nature of the political, cultural, and legal systems and institutions in which the feature exists as well.

My other concern is that in simply listing vulnerabilities we’re not really coming to grips with the risks inherent in each. But it is risk that enables us to assess cost, and we need cost to compare to benefits, and thereby arrive at a determination (this oversimplifies greatly, obviously, since factors being considered here are not necessarily commensurable).